Investigation Of Chemometrics Methods To Optimize Capillary Electrophoresis Separation And Extract Information

Capillary electrophoresis (CE) has been one of the most rapidly developed micro-separation analytical techniques in the latest twenty years. With the development of CE, chemometrics has been playing an important role in CE. In this thesis, new chemometrics methods such as wavelet tranform, genetic algorithm were deeply investigated to CE separation optimization and information extracting from CE data, which will contribute greatly to the theory and application of CE.In this thesis, a new method to determine the electroosmotic mobility in CE based on wavelet transform was developed. The rule of current change was studied during capillary electrophoresis separation process while the conductivity of the sample solution was different from that of the buffer. The time corresponding to sharp variation of current was detected using the derivative characteristic of spline wavelet transform, and thus the electroosmotic mobility was calculated. The relative errors compared with neutral marker method were all below 2.5%. A wavelet neural network was built based on wavelet transform and back propagation neural network. Combined with the demermination of nucleosides in urine samples by CE, it was used as a powerful pattern recognition tool to distinguish breast cancer patients from healthy persons. The recognition rate for the training set reached to 100% and above 96% of people in the predicting set were correctly classified, which was greatly higher than that obtained by principal component analysis. Compared with standard back propagation neural network, wavelet neural network had stronger abilities of information extraction and approximation.A muti-parameter optimization method for CE separation was developed based on genetic algorithm using real-valued encodings. Discrete wavelet neural network combined with orthogonal design was applied to model the multi-factor affectedseparation process. Using the modeled wavelet neural network as the target function, the genetic algorithm was used to search the optimal separation condition. The method was applied in the separation optimization of amino acids derivatized with 2- (9-carbazole) ethyl chloroformate by micellar electrokinetic chromatography, and the obtained value of normalized resolution product was increased by 12.5%. In addition, a methodology based on the combination of desirability function approach and genetic algorithm was applied in the multicriteria-decision-making simultaneous optimization of epinephrine enantiomers by capillary electrophoresis. The simultaneous optimization of the three goals for efficient, fast and high sensitivity separation has been achieved, and the value of total desirability function was increased by 10.4%. Considering the complexity of chiral CE separation mechanism, an multicriteria-decision-making optimization approach based on binary-coded GA, which was independent of the statistic model used, was proposed to CE separation of warfarin enantiomers.In separation of enantiomers by capillary zone electrophoresis (CZE), the concentration of chiral selector is a crucial variable. New equations to obtain the optimal concentration of chiral selector in buffer for the maximum resolution of the enantiomer separation by CZE were deduced with the consideration of both thermodynamics and kinetics characteristics of the migration of solutes. The organic solvent instead of water as separation medium has been an important approach to increasing separation selectivity in CE. The characteristics of the electroosmotic flow and the migration of acidic solute in non-aqueous capillary electrophoresis (NACE) were studied, and compared with theses in aqueous capillary electrophoresis. NACE was applied to the separations of nonsetroidal anti-inflammatory drugs and retionic acid isomer.